Global Group Testing and Screening With Dynamic Effects.

Area of Science:

• Biostatistics
• Bioinformatics
• Genomics

Background:

• Identifying outcome-related variables is crucial in biomedical research.
• Dynamic variable effects can complicate analysis and obscure important scientific mechanisms.
• Existing methods may not adequately account for time-varying covariate effects.

Purpose of the Study:

• To propose a novel model-free framework for testing and screening variables with dynamic effects on an outcome.
• To enable robust identification of dynamically associated variables and flexible group testing.
• To develop screening procedures for ultra-high dimensional data that ensure variable identification.

Main Methods:

• Utilizing a global perspective based on interval quantile independence.
• Developing a model-free testing strategy for dynamic variable effects.
• Extending the testing strategy to unconditional and conditional screening procedures.

Main Results:

• The proposed framework robustly identifies variables dynamically associated with outcomes.
• The method supports simultaneous evaluation of multiple continuous or discrete covariates (group testing).
• Screening procedures demonstrate the sure screening property in ultra-high dimensional settings.

Conclusions:

• The novel framework provides a powerful tool for analyzing dynamic variable effects in biomedical research.
• The methods show practical utility through simulations and real-world microarray data analysis.
• This approach enhances the accurate identification of key variables in complex biological systems.